Well, I spent my weekend digging in to this problem and I think I've found the root cause! An optimization in our scaling code was dropping samples. It only affects certain combinations of tracks / loops / samplerates / latencies so that explains why it was hard to reproduce.
Well, I spent my weekend digging in to this problem and I think I've found the root cause! An optimization in our scaling code was dropping samples. It only affects certain combinations of tracks / loops / samplerates / latencies so that explains why it was hard to reproduce.